How Power BI pseudo-consultants work
In market relations, the client and the consultant strive primarily for their own benefit: one — to buy a service cheaper, the other — to sell it more expensively.
Minimizing the costs of Power BI consultants is hidden in the pseudo advantages - the own solution out of the box. As a result, the system is adjusted by "adjusting" the client's data to a ready-made template.
As a result:
Analysis of data quality in the client's sources was not performed, recommendations for improving data purity were not provided.
Consultants have only a narrow range of information in their field of vision, the maximum potential of the data has not been revealed.
The template provides a set of indicators that are unfamiliar to the client, difficult to understand and not in demand within the company. The client simply does not understand what is really happening and does not realize that he is paying extra money.
When implementing a new system, pseudo-consultants often do not allow the client's company personnel to work, in order to further develop the system only with the involvement of the same consultants.
Also, we are often faced with the lack of technical knowledge of specialists implementing business analytics products:
Narrow knowledge of the Microsoft product line, as a result of incorrect selection of technology for the client's needs.
Lack of competence in the process of extracting, transforming and loading data (ETL) from sources. As a result, part of the data is missing in the final model.
Lack of competence in building a data warehouse for a business intelligence system, as a result, low speed and scalability of the solution.
Lack of competence in writing optimal formulas for calculation indicators, as a result of low speed of displaying information in boards.
By the nature of their activity, we often meet companies with Power BI implemented. Bright examples:
Horeca business - check data is taken from 1C, but there is a Front system. As a result of the source analysis, deviations from the primary source were found. As a result, the number of checks and all derived indicators is incorrect.
One of the clients has a "Data Lake" set up from Excel files for an analytical system. Preparation and uploading of files fell on the shoulders of the Customer. As a result, the rules of "integrity" of data were violated, in simple words - part of the data did not make it to the reporting.
At the retail client, the model is configured directly in the DBMS of the main transaction system. As a result, we received a double load on the database and, as a result, slow reports in Power BI and dissatisfied customers in the store.
The right approach to implementing a business intelligence product:
Study of the client's business model
Conducting meetings with key stakeholders to gather and analyze client requirements
Data analysis in company sources
Selection of subject areas, definition of business customers and pain points
Development of project passport
Approval of the model and technical task
Deployment of the site
Settings for downloading data from sources
Setting up data windows
Data visualization
Testing by the customer
Access settings
Training of key users
Preparation of documentation
Commissioning
A true architecture using the Microsoft stack:
Building a data warehouse - a separate database in MS SQL Server.
Configuring data loading using SSIS provides flexibility in the data extraction, transformation, and loading (ETL) process.
Building data models in the SSAS service, one of the main advantages is the ability to update model data in parts, a high rate of return. And don't forget to set up personal access to data through AD, which saves support time.
Report Server for Power BI is its own portal with an unlimited number of connections, it can also look at the world. This will allow you to configure the board for the owner in the mobile phone.
Excel experts have the ability to connect to a model in SSAS and build pivot tables based on model data. Also, if necessary, it is possible to organize the reverse recording of data from Excel to the data warehouse.
And we almost forgot, there is the ML (Machine learning) service. Feel free to use repository data and use Python or R packages.